|
PureBytes Links
Trading Reference Links
|
The problem with this is that if there's a statistically valid correlation - it
shouldn't exist only 180-190 days out (although perhaps 180-190 days might be the
strongest correlation). As with all robust systems - there should be a curve of
results - from good - to better - to best - not just a single "best" result smack
dab in the middle of some mediocre ---> terrible results. What's the correlation
100 days out - and 240 days out? If you're looking at garbage in those time frames
- all you're looking at is curve-fitting. Robyn
Ullrich Fischer wrote:
> He looked at the bond yields daily data and the daily gold price data and
> calculated correlation co-efficients between them using various time
> offsets in the bond data. He had a little table showing the correlation
> co-efficients for various days offset between the two data streams. It is
> a fairly straightforward process, once you have the bright idea to try
> it. I have no idea why Siatta thought to try doing this, but you can do it
> with a standard spreadsheet. The bottom line is he found a 65% correlation
> between the bond data and the gold data when comparing today's bond data
> with gold data from 180 to 190 days ago. That makes gold price a very
> significant (statistically speaking) predictor of bond yields approximately
> 6 months in the future. There is a discussion of this subject at
> http://www.equitytrader.com/ including a reference to Technical Analysis of
> Stocks & Commodities May '99 - Alex Siatta where Bollinger got the original
> concept. There is more info on the concept on www.BollingerBands.com
>
> -uf
|